Slashdot Mirror
IBM Watson To Battle Patent Trolls
MrSeb writes "IBM's Watson is made of many parts: speech recognition, natural language processing, machine learning, and data mining. All of these factors were perfectly combined to beat Ken Jennings in Jeopardy, and now each of these components are slowly finding their way into other applications. Health plan company WellPoint, for example, is using Watson to investigate patient records to improve diagnosis, and in a self-referential, possibly universe-destroying twist, IBM itself is using Watson to help sell Watson (and other IBM products) to other companies. Now, using Watson's data mining and natural language talents, IBM has created the Strategic IP Insight Platform, or SIIP, a tool that has already scanned millions of medical patents and journals for the sake of improving drug discovery — and in the future, it's easy to see how the same tool could be used to battle patent trolling, too."
..can someone send the patent office one of these machines for the love of god!
I'm afraid I can't let you patent rectangular shapes, Apple
I wonder if they used their own patent archives to train Watson to recognize junk patents?
It seems that the only person saying that Watson *could* be used "to battle patent trolls" is the last article's author. Nobody else has said that IBM or any customer using Watson is actually pursuing this use.
It has to be remembered that IBM is one of the biggest pro-software-patent lobby groups in the world.
In the US Bilski case, they submitted a brief saying that free software needed software patents!
http://en.swpat.org/wiki/Fake_representatives_of_free_software#IBM
I saw them personally in the EU lobbying from 2003-2005 where they pushed with all their might for software patents.
And then recently, when New Zealand announced it would legislate to clarify that software *isn't* patentable, who stepped in to kneel on the government? IBM (with MS).
http://yro.slashdot.org/story/10/06/23/0235248/new-zealand-u-turns-will-grant-software-patents
So, yeh, I'd be happy if all patent trolls disappeared tomorrow, but trolls aren't even the biggest problem, and the existence of the whole problem is in a large part due to IBM.
* http://en.swpat.org/wiki/More_than_trolls
* http://en.swpat.org/wiki/IBM
Expert in software patents or patent law? Contribute to the ESP wiki!
What is the business sense for IBM to do so, or the patent department ?
We as a society should insist the patent business gets cleaned up dramatically because innovation is getting bogged down.
Soon I would not be able to make a toast because it is somewhat rectangular with rounded edges and flat.
— and in the future, it's easy to see how the same tool could be used to battle patent trolling, too."
and it's also easy to see how the same tool could be used to automatically generate even more patents.
After all, since we've already seen that computers can randomly generate fake nonsensical Physics research papers and get them published in real Science Journals. We're not so far off that they'll be able to do the same with patent claims. It would be just like a turing test, but only easier since real patent legal language is already designed to obfuscate the obvious -- it would be easy to have a computer mimic it.
Train the Watson AI how to make executive and upper level management decisions. It would be possible to achieve the same results, but without the huge overhead of golden parachutes and high salaries. What remains could be summed up as profit.
For IBM, there's plenty of incentive for them to do find grounds for challenging every patent that isn't held by IBM (though little incentive for them to reveal those grounds until the patent is used in way which hurts IBMs business.)
For the patent office, determining what is and isn't patentable under the law and only approving applications for the former category is their job, so an automated tool that makes it easier to make that determination correctly would be in their interest.
Watson never did speech recognition. It was delivered the questions electronically.
One of these machines falls into the wrong hands and patent trolls use it with a small bit of programming to create patent applications for them; and by that I mean... machine generated patents
In other words... entirely nonsensical patents for technology that has never been actually used, and might not actually work, but an extremely massive number of eloquently written machine-generated patents covering every conceivable problem with weird vague claims, with problem sets, and concepts for machine-generated patent claims obtained from automatic mining of past patent language, weblogs, etc...
The invention might not be real by any stretch of the imagination, but the awarded nonsensical patent might be vague enough to actually sue over a technology actually invented in the future
In other words... the ultimate patent troll is a computer AI that generates convincing applications in massive numbers without actually inventing anything.